Research On Intelligent Analysis And Visual Detection Technology Based On Images Of Hainan Gibbons

The Hainan gibbon listed as critically endangered is an important symbol of the originality and integrity of the rainforest.At present,the number of Hainan gibbons is mainly counted manually and through passive acoustic detection.However,due to people poaching in the early years,Hainan gibbons are wary of people.Besides,the jungle is dark,wet,and slippery,which is not conducive to researchers’ work and makes the accuracy greatly reduced.In this thesis,we use ArcGIS to intelligently analyze Hainan gibbon video data,combine deep learning to study visual detection and design Hainan gibbon detection system with edge computing.The main work is as follows.1.Construct the activity pattern and activity trajectory map of Hainan gibbons.In response to the problem that the existing wildlife cameras deployed in the Bawangling area often fail to capture images of Hainan gibbons,the elevation information of existing surveillance videos is classified.It uses the digital elevation model of ArcGIS10.5 based on the idea of weighted averaging.And summarizes the geographical activity trajectory information about Hainan gibbons in different temperatures,seasons,and time nodes.It provides the basis for optimizing the location deployment of the wildlife camera system at a later stage.2.Propose an image filter(IF-YOLO)based object detection method for Hainan gibbons.The unique cloud forest environment of the Hainan tropical rainforest affects object detection.A pre-convolutional neural network is used to processing to make image enhancement and suppress the interference of weather information.Two networks are combined for end-to-end detection.Experimental results show that the method can improve the model’s performance in detecting Hainan gibbons in low-quality images under rain and fog weather.3.Design two active-passive Hainan gibbon detection systems and build the software interface of the Hainan gibbon detection system.In order to realize the global capture of Hainan gibbon images,this thesis designs mobile monitoring platforms based on two ways of tiller head and all-round platform on Raspberry Pi to realize the edge detection of Hainan gibbon.And build the software interface to realize user interaction to detect Hainan gibbon,the experimental results provide new ideas for the improvement and development of current wildlife cameras.In summary,this thesis proposes a complete image solution for the Hainan gibbon from camera deployment and object detection to edge detection,which provides theoretical support for Hainan gibbon conservation.